Comparison of M5 Model Tree and Artificial Neural Network’s Methodologies in Modelling Daily Reference Evapotranspiration from NOAA Satellite Images

The objective of this study was to compare feed-forward artificial neural network (ANN) and M5 model tree for estimating reference evapotranspiration (ET₀) only on the basis of the remote sensing based surface temperature (T_s) data. The input variables for these models were the daytime surface temperature at the cold pixel obtained from the AVHRR/NOAA sensor and extraterrestrial radiation (R_a). The study has been carried out in five irrigated units that cultivate sugar cane, which located in the Khuzestan plain in the southwest of Iran. A total of 663 images of NOAA–AVHRR level 1b during the period 1999–2009, covering the area of this study were collected from the Satellite Active Archive of NOAA. The FAO-56 Penman–Monteith model was used as a reference model for assessing the performance of the two above approaches. The study demonstrated that modelling of ET₀ through the use of M5 model tree gave better estimates than the ANN technique. However, differences with the ANN model are small. Root mean square error and R² for the comparison between reference and estimated ET₀ for the tested data set using the proposed M5 model are 13.7 % and 0.96, respectively. For the ANN model these values are 14.3 % and 0.95, respectively.